All Data are Useful, but not All Data are Used! What'S Going on Here?

Issues limiting the use of modern data streams are described and possible approaches are outlined. Data assimilation system constraints and limitations are listed; a taxonomy of the various errors involved is given; and the difficulty of showing positive impacts for new or currently unused data is discussed. Improvements may come from creative, parsimonious representations of the principal errors; adaptively selecting or pre-processing the data; and questioning the basic assumptions made in current operational data assimilation systems. Examples given from the perspective of satellite scatterometry and IR retrievals illustrate problems of general interest.